An AI decision is also a process decision.
Buy a tool, build one, or automate a step, and the way the work gets done changes with it.
I'm Guy Pavlov . I write short notes on where AI is worth using, what it costs, what to check before a tool sees client data, and how to tell whether it helped.
If AI doesn't seem relevant to your business, test the tasks your team repeats. Four questions show where to try it and when to leave it alone.
What to work out next
Choosing tools
Build or buy? How smaller firms should choose AI tools
How a smaller firm chooses between buying an AI tool, building one, ordinary automation, changing the process, or leaving it alone, starting from the problem.
AI or automation
AI vs automation: how to choose for everyday work
If you can write the rule down as “when this happens, do that”, ordinary automation does the job. If a step needs judgement, use AI and have a person check it.
Which AI to pay for
Which AI is best for a small business? Test it on your team's work
Which AI a small business should pay for: try two business plans on work your team repeats, then check how each reaches your apps and what it costs.
Choosing a plan
ChatGPT Business vs Plus: which plan to buy
ChatGPT Business vs Plus for small firms: what each costs, who controls access, how your data is used, and a checklist for choosing between them.
Costs
What AI adoption costs a small firm, and what it pays back
What AI adoption costs a small firm beyond the subscription, what AI consultants charge, and how to count hours recovered apart from cash saved.
Client data
Client data and AI tools: six checks before you sign up
The six checks I make before an AI tool touches client data, from approved tools and retention to access, disclosure and the EU AI rules.
AI use policy
AI use policy template for a small business
A free one-page AI use policy template in Word and Markdown: approved tools, information that never goes in, who checks the output, new tools and review dates.
Measuring results
How to tell whether an AI tool is actually helping
How to tell whether an AI tool is helping: write down a baseline, agree the measure, check it once people use the workflow, and report each gain separately.
Training your team
How to train employees to use AI on one real task
A one-hour session a manager can run: pick one task, prepare sample inputs, catch a flawed AI answer, check the work, and follow up a week later.
By kind of firm
Engineering teams
How a stack review cut a SaaS company's AI costs
How a review of a SaaS company's coding tools, workflows and model spend cut its AI costs by about 80%, and the engineering measures I track afterwards.
Client-service firms
Start with the work you redo for every client
The audit or analysis your team redoes for each client is often the best first task for AI. Two examples from agency work, with the hours saved.
Bring one task to a call.
Pick the one that takes your team too much time, and we'll work out whether a review makes sense. 30 minutes, free, no preparation needed. I'll tell you on the call if there's no opportunity.

Guy Pavlov The call is with me.